Multi-Scale RAFT: Combining Hierarchical Concepts for Learning-based Optical FLow Estimation
This work addresses optical flow estimation for computer vision applications, presenting an incremental improvement by integrating multi-scale ideas into an existing successful method.
The paper tackles the problem of optical flow estimation by proposing a multi-scale neural network based on RAFT, which combines hierarchical concepts like a coarse-to-fine architecture and multi-scale features. The result shows substantial improvements over RAFT and achieves state-of-the-art results on MPI Sintel and KITTI datasets, particularly in non-occluded regions.
Many classical and learning-based optical flow methods rely on hierarchical concepts to improve both accuracy and robustness. However, one of the currently most successful approaches -- RAFT -- hardly exploits such concepts. In this work, we show that multi-scale ideas are still valuable. More precisely, using RAFT as a baseline, we propose a novel multi-scale neural network that combines several hierarchical concepts within a single estimation framework. These concepts include (i) a partially shared coarse-to-fine architecture, (ii) multi-scale features, (iii) a hierarchical cost volume and (iv) a multi-scale multi-iteration loss. Experiments on MPI Sintel and KITTI clearly demonstrate the benefits of our approach. They show not only substantial improvements compared to RAFT, but also state-of-the-art results -- in particular in non-occluded regions. Code will be available at https://github.com/cv-stuttgart/MS_RAFT.